Why can’t utils.SuspendWorkflow() pause a workflowrun that has a step-group?

[moruikang]

Describe the bug

I found that in the previous version of the SDK(vela cli 1.7), SuspendWorkflow() would only set run.Status.Suspend = true.

However, the new version of the SDK(vela 1.9) will set sub-step’s sub Phase to: v1alpha1.WorkflowStepPhaseSuspending. This could lead to the workflowrun Suspending state finally being set to state=Executing in Reconcile().

please tell me why? and how can i use utils.SuspendWorkflow() to stop my workflowrun?

To Reproduce

my demo workflow yaml:

apiVersion: core.oam.dev/v1alpha1
kind: WorkflowRun
metadata:
  name: dependency-ng
  namespace: default
spec:
  mode:
    steps: DAG
  workflowSpec:
    steps:
      - name: step1
        type: apply-deployment
        dependsOn:
          - step2
          - step3
        properties:
          image: 10.113.xx.x:31104/library/nginx:v1.23.2
      - name: step2
        type: apply-deployment
        properties:
          image: 10.113.xx.x:31104/library/nginx:v1.23.2
      - name: step3
        type: apply-deployment
        properties:
          image: 10.113.xx.x:31104/library/nginx:v1.23.2
      - name: step-gg
        type: step-group
        subSteps:
          - name: apply-sub-step1
            type: apply-deployment
            properties:
              image: 10.113.xx.x:31104/library/nginx:v1.23.2
          - name: apply-sub-step2
            type: apply-deployment
            properties:
              image: nginx:v1.23.2

The following is the official source code call (workflow 0.6)：

clientN := mgr.GetClient()

run := &wfv1alpha1.WorkflowRun{}
ctx := context.Background()
key := types.NamespacedName{Namespace: "default", Name: "dependency-ng"}
err = clientN.Get(ctx, key, run)
if err != nil {
    fmt.Println("err1")
    panic(err)
}

err = utils.SuspendWorkflow(ctx, clientN, run, "")

if err != nil {
    fmt.Println("err2")
    panic(err)
}

workflowrun cant not suspend

$ kubectl get workflowruns.core.oam.dev
dependency-ng   executing   23m

The following is my custom call based on the workflow source code of the previous version(workflow 0.4-)：

	//Manually try to pause the task
	stepName := ""
	run.Status.Suspend = true
	steps := run.Status.Steps
	found := stepName == ""

	for i, step := range steps {
		if step.Phase == wfv1alpha1.WorkflowStepPhaseRunning {
			if stepName == "" {
				utils.OperateSteps(steps, i, -1, wfv1alpha1.WorkflowStepPhaseRunning)
			} else if stepName == step.Name {
				utils.OperateSteps(steps, i, -1, wfv1alpha1.WorkflowStepPhaseRunning)
				found = true
				break
			}
		}
		for j, sub := range step.SubStepsStatus {
			if sub.Phase == wfv1alpha1.WorkflowStepPhaseRunning {
				if stepName == "" {
					utils.OperateSteps(steps, i, j, wfv1alpha1.WorkflowStepPhaseRunning)
				} else if stepName == sub.Name {
					utils.OperateSteps(steps, i, j, wfv1alpha1.WorkflowStepPhaseRunning)
					found = true
					break
				}
			}
		}
	}
    if err := retry.RetryOnConflict(retry.DefaultBackoff, func() error {
		return clientN.Status().Patch(ctx, run, client.Merge)
	}); err != nil {
		fmt.Println(err)
	}

workflowrun Suspend work

$ kubectl get workflowruns.core.oam.dev
dependency-ng   suspending   23m

Expected behavior

Screenshots

Workflow Version

v0.6.0

Cluster information

Additional context

[FogDong]

Can you provide more detail of this?

[moruikang]

Can you provide more detail of this?

As above, I have provided